Margulis–Platonov: local origins of abstract normal subgroups

← All digests

Family 018 · Number theory
Importance 4/5 · Provisional editorial rating
Sources consulted 2026-10-10

1. The question

A normal subgroup is the kernel of a possible quotient: it describes which elements become indistinguishable after passing to a simpler group. For matrix groups with entries in a number field, one can construct quotients by examining the entries locally, using completions of the field and congruence conditions.

Could there also be abstract normal subgroups that have no such local explanation? The word “abstract” is important. We do not assume that a subgroup is closed, continuous, finitely generated or initially related to congruences.

The Margulis–Platonov conjecture predicts that, apart from subgroups in the center, every normal subgroup of the rational points of a suitable simple algebraic group is detected by finitely many compact local groups. Family 018 claims this classification over number fields and over global function fields, including characteristic two.

2. Background and essential definitions

Global fields and their completions

A global field is either a finite extension of $\mathbb Q$, called a number field, or a finite extension of $\mathbb F_q(t)$, called a global function field. A place $v$ specifies an absolute value, and completing the field gives $k_v$.

For $\mathbb Q$, the familiar real absolute value gives $\mathbb R$, while each prime $p$ gives the field $\mathbb Q_p$. These latter, nonarchimedean completions measure divisibility rather than ordinary size. Function-field places are all nonarchimedean.

An algebraic group $G$ is a group defined by polynomial equations. Its rational points $G(k)$ consist of solutions with coordinates in $k$. The standard example is $\mathrm{SL}_n$, the determinant-one matrices.

The hypotheses are that $G$ is absolutely almost simple and simply connected. Absolutely almost simple means that, even over an algebraic closure, it has no proper positive-dimensional connected normal algebraic subgroup. Simply connected has its algebraic-group meaning: there is no nontrivial central isogeny onto it from another connected semisimple group. It does not assert topological simple connectedness of every group $G(k_v)$.

The compact local factors

The $k_v$-rank of $G$ is the dimension of its largest split torus over $k_v$. A split torus looks like a product of multiplicative groups. If this rank is zero, the group is anisotropic at $v$.

For the nonarchimedean places in this setting, anisotropy corresponds to compactness of $G(k_v)$. Compact groups can have open normal subgroups and finite quotients. They are therefore natural sources of finite quotients of $G(k)$.

Let

$$\mathcal A=\{v\text{ nonarchimedean}:\operatorname{rank}_{k_v}G=0\}, \qquad H_{\mathcal A}=\prod_{v\in\mathcal A}G(k_v).$$

The set $\mathcal A$ is finite. The diagonal map $\delta_{\mathcal A}:G(k)\to H_{\mathcal A}$ regards the same rational element in each completion. The product has its usual local topology.

The center $Z(G(k))$ consists of elements commuting with every element of $G(k)$. A subgroup is noncentral if it is not contained in this center. If $W$ is open and normal in $H_{\mathcal A}$, its inverse image is an abstract normal subgroup of $G(k)$. The conjecture says that this construction accounts for all noncentral normal subgroups.

3. Development history and previous partial results

The problem grew from Kneser’s simplicity question for norm-one groups of quaternion algebras and Platonov’s extension to simply connected simple groups. Margulis formulated the description using anisotropic local factors. His normal-subgroup results supplied a crucial finite-index reduction, but finite index alone did not identify which local topology detects the subgroup.

The isotropic case was settled through the global Kneser–Tits theorem and simplicity results based on unipotent generation. The anisotropic cases require different tools because those unipotent directions may be absent over the ground field.

For inner forms of type $A$, the groups are closely related to norm-one elements of division algebras. Work by Tomanov, Platonov–Rapinchuk, Raghunathan and Rapinchuk–Potapchik developed reductions to finite quotients. Segev and Segev–Seitz supplied finite-simple-group obstructions. Rapinchuk–Segev–Seitz then proved the stronger statement that every finite quotient of the multiplicative group of a finite-dimensional division algebra is solvable, and established Margulis–Platonov for inner type $A$ over global fields. Their primary paper records this result.

Other classical and low-index unitary cases were known, as were substantial quadratic-splitting cases through Chernousov’s work. The number-field manuscript identifies the remaining groups as anisotropic outer type $A$, trialitarian type $D_4$, and type $E_6$. Over function fields, Harder’s structural results reduce the unresolved anisotropic case to outer type $A$, represented by special unitary groups.

The conjecture also matters to the congruence subgroup problem, which asks whether finite-index subgroups of arithmetic groups come from congruence conditions. Prasad and Rapinchuk’s survey explains this relationship. It is an input to that broader problem, not an automatic solution to all its cases.

M2 develops the number-field argument. M1 adapts its structure to function fields, replacing steps that depend on characteristic zero and treating characteristic two separately where necessary.

4. What the new manuscripts claim

The normal-subgroup theorem over all global fields

M2 claims that, over every number field $k$, every noncentral abstract normal subgroup $N\triangleleft G(k)$ has the form

$$\boxed{N=\delta_{\mathcal A}^{-1}(W)}$$

for some open normal subgroup $W\triangleleft H_{\mathcal A}$. Only nonarchimedean anisotropic places enter the product. Compact real factors are not added to this formula.

M1 claims exactly the corresponding description for every global function field and every positive characteristic. It includes groups anisotropic over the global field and does not exclude characteristic two or power exponents divisible by the characteristic.

If $\mathcal A$ is empty, the product is the trivial group. The conclusion then reads

$$N=G(k).$$

Thus there are no proper noncentral normal subgroups in that case. This is an exact abstract equality, stronger than saying that a subgroup is dense or merely has finite index.

In general, an open subgroup of the finite compact product has finite index, so the theorem explains the finite-index conclusion through a particular local source. It also means that every finite quotient of $G(k)$ is continuous for the topology induced by these anisotropic local factors: its finite-index kernel is noncentral and hence arises from an open local subgroup.

A carefully delimited congruence consequence

M2 applies the normal-subgroup theorem to Rapinchuk’s positive-density congruence theorem. Let $M/k$ be the minimal Galois extension over which $G$ becomes an inner form of the split group. Let $\operatorname{Spl}(M/k)$ be the nonarchimedean places splitting completely in $M$.

Suppose a set of places $S$ contains every archimedean place, contains no nonarchimedean anisotropic place, and satisfies

$$d_k\bigl(S\cap\operatorname{Spl}(M/k)\bigr)>0,$$

where the indicated Dirichlet density exists. Dirichlet density measures the asymptotic proportion of primes through their weighted prime sums. Under these hypotheses M2 claims that the $S$-congruence kernel is trivial.

Equivalently, in a fixed faithful matrix representation, every finite-index normal subgroup of $G(\mathcal O(S))$ contains a principal congruence subgroup of nonzero ideal level. Here $\mathcal O(S)$ consists of elements integral outside $S$; a principal congruence subgroup is the kernel of reduction modulo an ideal. The conclusion retains the positive-density and anisotropic-place restrictions. It is not the unrestricted finite-$S$ congruence conjecture.

5. Why the advance matters

Provisional importance: 4/5. Taken as stated, the family completes a major structural classification across all global fields. It says that an abstract quotient cannot hide information invisible to the designated local groups.

This is a strong connection between algebra and topology. The subgroup is introduced without continuity, yet the theorem forces a local topological description. The anisotropic factors are precisely where compactness permits finite information to survive.

The result also supplies a missing hypothesis in existing congruence theorems and unifies remaining unitary and exceptional-group cases. Its impact is broad within arithmetic groups, while the separate requirements of the full congruence subgroup problem remain significant.

6. The main idea

Isolate what local closure might miss

Starting with a finite quotient, choose a positive integer $e$ annihilating it and form the subgroup

$$R=\langle g^e:g\in G(k)\rangle.$$

This is the group generated by all $e$-th powers, not merely the set of powers. Let

$$P_0=\overline{\delta_{\mathcal A}(R)},\qquad V_0=\delta_{\mathcal A}^{-1}(P_0),\qquad V=V_0/R.$$

The finite group $V$ measures a potential discrepancy: a rational point might lie in the local closure of the power subgroup without lying in the abstract subgroup itself. The central task is to prove $V=1$.

Ordinary approximation only gets close locally. The papers develop approximation within an exact prescribed coset of $R$, while simultaneously controlling finitely many places and avoiding unwanted reductions at the others. Products of conjugated torus powers retain the coset condition. A separate norm-torus calculation turns compatible local norm equations into exact rational solutions.

Exclude two kinds of finite obstruction

For an odd-degree division algebra with a unitary involution, a nontrivial discrepancy produces either a finite abelian character or a finite nonabelian simple quotient of the full unitary group. Passing from the special unitary subgroup to this full group requires commuting approximation and a central-extension argument; it is not automatic.

Finite abelian characters are handled by exact algebraic difference identities and elementary transformations, using the already known inner-type-$A$ theorem.

A nonabelian simple quotient can be viewed as assigning finitely many colors to rational unitary elements. The color assignment is not assumed continuous. The papers construct an invariant probability law on finite color patterns that retains uniform distributions over all colors. This is essential: naive averaging could lose some of them.

Recurrence then forces membership in a carefully chosen proper subset of colors to be invariant under more transformations than is possible. Finite-simple-group structure supplies the obstruction, and generated rotations ultimately force that subset to be either empty or the entire group, contradicting its choice.

Adapt the argument to characteristic and group type

The number-field proof uses a multidimensional recurrence theorem of Furstenberg and Katznelson. In positive characteristic, additive groups have finite exponent, so the same integer-dilation argument does not apply. M1 replaces it with finite additive Fourier analysis. Odd characteristic uses a quadratic parametrization; characteristic two uses additive polynomials and coordinates over power subfields.

Another issue arises when the characteristic divides $e$: taking $e$-th powers need not be locally invertible. M1 separates the inseparable part and controls local power subgroups directly.

Induction then handles the remaining unitary degrees through smaller centralizers, quaternion factors and simultaneous norm equations. Degree four has a separate determinant correction. M2 additionally treats the remaining $D_4$ and $E_6$ groups by factorization into subgroups whose normal-subgroup behavior is already controlled. Finally the argument returns from the power subgroup to the original $N$.

7. Scope, limitations, and open questions

The groups are absolutely almost simple and algebraically simply connected. Dropping those assumptions can change both the available quotients and the appropriate statement. Central subgroups are explicitly outside the classification of noncentral ones.

The theorem concerns $G(k)$, the full rational-point group. An arithmetic subgroup such as $G(\mathbb Z)$ is a different object and can have different normal-subgroup behavior. One should not replace it by $G(k)$ in either direction without the relevant arithmetic argument.

The congruence consequence is limited to the stated positive-density range. It does not settle Serre’s finite-$S$ conjecture, arbitrary infinite sets $S$, or all questions about the size and centrality of congruence kernels. Nor does the classification provide an explicit algorithm listing every open normal subgroup of each compact local factor.

8. Sources and further reading

Theorem 1.1 of M1 gives the function-field statement, including characteristic two. Theorem 1.1 of M2 gives the number-field statement; Corollary 1.2 gives the positive-density congruence consequence. Both introductions explain the finite discrepancy group and distinguish known type-$A$ inputs from the remaining cases.

The Rapinchuk–Segev–Seitz paper and Prasad–Rapinchuk survey linked in section 3 provide primary background. Rapinchuk’s positive-density theorem is the separate input used for M2’s corollary. Optional background on normal subgroups and local fields can help connect the abstract and topological sides.

Source manuscripts

• M1. OpenAI, The Margulis–Platonov conjecture over global function fields (October 5, 2026).
• M2. OpenAI, The Margulis–Platonov conjecture over number fields (September 23, 2026).

Sources consulted on 10 October 2026 at repository revision fd4aeeb. Download the manuscript citations.